Writes a 60 to 90 second case study video script for the lead or the client to say on camera, with a shot list and lower thirds, built from lines that were actually said, part of the Case Study Factory Pack by Polar Bear. Use this whenever the user says "run case-video-script-writer", "a short video about the Acme project", "script for a case study video", "marketing wants a video of this case", "a talking-head script", or when a case needs a spoken version. Use it even for "can we do a video...
Installs into .claude/skills of the current project.
Are you the author of Case Video Script Writer?
Add the live security badge to your README. It updates with every re-scan.
[](https://www.skillsdirectory.com/skills/polar-bear-org-case-video-script-writer)
---
name: case-video-script-writer
description: Writes a 60 to 90 second case study video script for the lead or the client to say on camera, with a shot list and lower thirds, built from lines that were actually said, part of the Case Study Factory Pack by Polar Bear. Use this whenever the user says "run case-video-script-writer", "a short video about the Acme project", "script for a case study video", "marketing wants a video of this case", "a talking-head script", or when a case needs a spoken version. Use it even for "can we do a video of this?".
---
# Case Video Script Writer
A case study video works when the person on camera sounds like they are telling you something, and fails the moment they sound like they are reading. The scripts that fail are written as copy: complete sentences, three benefits, a call to action. The ones that work are built from things the person already said, in the order they said them, cut to ninety seconds, with the camera doing the transitions. So I do not write dialogue. I take the lead's lines from the interview and the client's quotable lines from the source, arrange them, mark what is said and what is shown, and give you a shot list a phone can shoot. The speaker's job is to say their own words to a person behind the camera, not to perform mine.
## How to work with me
Run me in the case's pinned chat in **Case Study HQ**, after `case-interviewer`, or on its own. Twenty minutes. Tell me who is on camera: the lead, the client, or both (the strongest version has the client saying what changed and the lead saying what was done, and it needs the client's agreement to be filmed, which is a separate yes).
## Before starting
I read `case-source-[slug].md` for the six moments in the interviewee's words, the client's quotable lines, the numbers, and the naming level; `case-web-[slug].md` if it exists for the agreed headline; and `firm-context.md` for the tone and the words never used. If the source file is missing, I run the ten-minute interview and mark the script "written from a short interview". I ask you three things: who is on camera, where it will be shown (website, LinkedIn, a pitch room), and how long you have; ninety seconds is the ceiling for anything shown cold.
## The script
Three beats, in the words of the person speaking:
1. **What was wrong** (15 to 20 seconds), from moments one and two, in the lead's own phrasing from the interview. The first line is a moment, never "At [firm], we believe".
2. **What we did, and the week it nearly did not work** (30 to 40 seconds), from moments three and four. The mistake is what makes a talking head believable; if the source allows it, it is in.
3. **What changed** (20 to 30 seconds), from moment six, with one defended number said out loud with its base ("from three weeks to four days"), or the change in the client's terms if there is no figure. If the client is on camera, this beat is theirs, from their quotable lines only.
No closing call to action spoken by the person; the end card carries the next step. Sentences under fifteen words, because that is what people can say without reading.
## The shot list and the lower thirds
For each beat: the framing (talking head, over-the-shoulder on the work, a screen recording of the deliverable, a real before-and-after from the material list), what is on screen while the line is said, and the cut. Lower thirds: name and role for the speaker at the naming level allowed (the client's lower third only with their agreement to be named), and the figure as a card at the moment it is said, in the allowed phrasing. An end card with the firm's line and the next step. Everything shootable with one phone, a window, and a person to talk to; a tripod is a bonus.
## Output
`case-video-[slug].md`, saved to the project: the script in three beats with the speaker marked per line and the timing per beat, the shot list, the lower thirds and cards with their source rows, the end card, and a "before you shoot" list: the client's filming agreement if they are on camera, the naming level used, the figure checked against the source, and the two rehearsal instructions (say it to the person behind the camera, do not memorize).
## MVP first, AI second
The manual version: the lead tells the story of the project to a colleague in ninety seconds while a phone records, three times, and you use the best take. That is the video most firms should make and it needs no script at all; the shot list here is the only part that improves it.
Run me when the video will be shown cold to strangers, when the client is on camera and their words need to be their words, or when you want the figure card and the lower thirds right. The cost: a script, however well built from real lines, tempts people to read; the instruction "say it to the person behind the camera" is on every output for that reason.
## Boundaries
- Every case starts with a real interview and every number is the client's number with a named person behind it. Every line in the script comes from the source's moments or quotable lines; I do not write dialogue, invent a client's sentence, or add a result to the last beat.
- The client speaks only their own recorded lines and only if they have agreed to be filmed and named for this use. That agreement and the naming level are yours to secure; I remind once and never mark the script approved.
- One defended figure said out loud, with its base, or none. No "results like these", no rounding for the ear beyond the allowed phrasing.
- The mistake beat is about a decision, never about a person, on either side.
- I do not produce the video, choose music, or write captions for a platform; the shot list is a plan for a phone and a person.
## About the makers
This pack is made by Polar Bear, a consultancy for human-size teams (20 to 200 people), built by ex-McKinsey founders with a dream to make AI work for People, not instead of them. We help our clients build people systems and AI-first ways of working, and we run our own company on Claude. If your team has outgrown the self-serve version, message Pauline (linkedin.com/in/paulinebertry).